TL;DR
This paper introduces a novel text-based method to track the "vogue" of ideas in American sociology, moving beyond traditional citation and co-authorship metrics. They found that research trends are driven by applied topics and shaped by regional and institutional factors, rather than a simple top-down diffusion from elite institutions.
What Problem Does This Paper Address?
Understanding how ideas gain traction and spread within academic disciplines is crucial for grasping the evolution of knowledge. Traditional methods for studying this, like citation analysis or co-authorship networks, offer a macro-level view but can miss the nuances of how specific concepts emerge, circulate, and become influential. This paper aims to bridge this gap by developing a more fine-grained, text-based approach to identify and track "vogue" – the rise and fall of ideas – within American sociology, exploring the underlying social and institutional dynamics.
Methods
The authors introduce a novel method that leverages semantic network analysis and natural language processing (NLP) on a large corpus of sociology dissertations. They construct semantic networks by analyzing co-occurring word pairs within dissertation abstracts. A "vogue" term pair is identified when it moves from the periphery to the "backbone" (most significant connections) of the semantic network over time, indicating increased prominence. They also analyze the "foundational" terms that remain consistently on the backbone across different time periods. This approach allows them to track the emergence, diffusion, and decline of research trends.
Datasets
The primary dataset consists of 4,747 sociology doctoral dissertations completed between 2011 and 2020 from 114 PhD-granting institutions in the United States. They also utilized journal reputation rankings from "All Our Ideas" to assess the influence of different journals.
Key Insights and Ideas
- Applied Topics Drive Trends: Surprisingly, applied research areas like crime and health, often perceived as less prestigious, are the primary engines behind emerging and diffusing research trends in sociology.
- Beyond Elite Dominance: Knowledge production isn't a simple trickle-down from top-tier institutions. Instead, it's a more complex, segmented landscape influenced by a interplay of factors.
- Regional and Institutional Fingerprints: Geographical location and institutional classification (public/private) play a significant role in shaping how research trends are adopted and spread, suggesting regional clustering and shared characteristics are key.
- Divergent Niches, Not Just Mimicry: Lower-status institutions don't just copy elite trends; they tend to cultivate distinct research niches that differ from those dominating core institutions.
- Vogue vs. Foundation: The study distinguishes between "vogue" terms (emerging, timely topics) and "foundation" terms (broader, enduring concepts), revealing different roles in the discipline's intellectual structure.
Key Results
- The study identified "vogue" term pairs that significantly increased in co-usage frequency from the first half of the decade (2011-2015) to the second half (2016-2020).
- Journals focusing on empirical social issues (e.g., Journal of Health and Social Behavior, Demography) showed the highest overlap with vogue terms, indicating their trendsetting influence, rather than the most prestigious theoretical journals.
- Public institutions in the same geographic region were more likely to adopt each other's trends, while similarly ranked public institutions showed less adoption, possibly due to competition.
- Private institutions, however, showed increased adoption among similarly ranked peers, suggesting a different dynamic of exchange.
- Institutions were more likely to adopt trends that aligned with their existing intellectual trajectories (research fit).
Limitations and Future Work
- Textual Data Limitations: Reliance on dissertation abstracts might overlook unpublished work, informal discussions, and non-textual forms of intellectual engagement, potentially missing crucial diffusion pathways.
- Temporal Granularity: The study splits the decade into two periods, which might not capture rapid shifts or long incubation periods for certain ideas. Future research could extend the analysis over longer time spans.
- Randomness in Term Usage: Year-to-year variations in term usage by graduates introduce some randomness.
- Future Directions: The authors suggest exploring the impact of government funding policies on knowledge production and diffusion, and employing more advanced techniques like hypergraph modeling for semantic network analysis.
Summary updated